The first decade of coherent optics development focused on pushing the performance envelope, driving higher capacity per wavelength and longer optical reach. As coherent optics enter their second decade of development, the traditional metrics used to compare coherent capabilities, such as the "headline capacity" of maximum wavelength speed, may no longer define the best solution for a given application.
Indeed, the one-size-fits-all approach will give way to specialists who tailor products for new customer use cases. Beginning this year, a wide range of products using 7-nanometer coherent Digital Signal Processors (DSPs) will enter the market, helping expand the use of coherent optics across a wide range of network applications.
As coherent optics development moves away from a one-size-fits-all approach, the market dynamics will continue to favor suppliers that are vertically integrated around coherent DSP technology. This will undoubtedly disrupt the traditional competitive ecosystem between optical components, modules and system suppliers.
The bifurcation of coherent optics development will have some seeking to push the next generation of high-performance coherent optics towards ever-greater capacity-reach capabilities. Others will opt to chase low-power applications.
In the max-reach camp, developers will use high-speed DSPs and electro-optics operating at the highest baud rates possible, typically between 90-100 Gbaud. These devices will implement high-gain Forward Error Correction (FEC) to correct transmission errors, Nyquist shaping to optimize spectral use, and Probabilistic Constellation Shaping (PCS) to dynamically adjust modulation format, all with the goal of maximizing capacity into a given wavelength spectrum (known as spectral efficiency), while simultaneously pushing optical reach to the longest distances possible.
This will push coherent performance to within 1-2dB of theoretical limits, effectively defined by Shannon's Law. While the use of smaller silicon CMOS node geometries will reduce DSP power, high-performance coherent optics will nevertheless necessitate implementation in chassis-based optical transport systems. These next-gen coherent products represent a continuation of the industry's traditional development focus, and key suppliers will remain the large, vertically integrated coherent optical systems vendors: Ciena, Huawei, Infinera and Nokia -- all vendors that have, or are expected to announce, new products in this area.
Acacia has made a recent entry into this space, predicated on the prior generation of 16nm DSPs operating up to 600 Gbit/s, so as to offer a merchant vendor alternative to systems vendors lacking in-house coherent technology. It remains to be seen if this strategic focus will continue following Acacia's expected acquisition by Cisco.
At the same time, coherent optics optimized for low power, rather than maximum capacity-reach, will enable a new class of pluggable coherent transceivers, and extend coherent use cases into metro and access applications. Also leveraging 7nm DSP technology, combined with highly integrated, miniaturized transmitter-receiver optical sub-assemblies (TROSAs) operating in the 60-70Gbaud range, these power-optimized pluggable coherent optics will support maximum wavelength speeds up to 400Gbit/s, and maximum optical reaches of several hundreds of kilometers.
The concept of power-optimized or "compact" coherent optics is not new; however this next generation of low power coherent optics will now enable implementation in datacom-centric transceiver formats such as QSFP-DD and OSFP. This will allow packet switches to interchangeably utilize either short-reach datacom optics for intra-datacenter (DC) connections or coherent WDM optics for inter-DC links, using the same ports. This removes the historical trade-off of having to build dedicated ports for WDM optics in a different form factor, and the resulting loss in switch I/O density.
Low-power pluggable coherent optics will enable upgrade of inter-DC connections to 400 Gigabit Ethernet (400GE) speeds, and have gotten considerable traction from the network operator community. The 400G ZR specification codified by the Optical Internetworking Forum (OIF) will enable multivendor product interoperability for point-point links up to 120km using optical amplification, and most 400G ZR vendors are planning module variants that push optical reach up to several hundreds of kilometers, under the commonly used but loosely defined moniker, 400G ZR+.
The first shipments of next-gen 400G coherent transceivers are expected from Acacia and Inphi. Not far behind will be similar products from component vendors II-VI, Lumentum, Neophotonics and others, leveraging their product strength in TROSAs combined with externally sourced DSPs from NEL or Inphi.
Also entering the coherent transceiver market will be system vendors leveraging their vertical integration in coherent optics. Both Ciena and Infinera having already announced plans in this space. This will create competition between companies which were formerly suppliers and customers. If all that wasn't disruptive enough, system vendors lacking viable in-house coherent optics capabilities will buy their way in, as Cisco is doing with its offer to acquire Acacia for $2.6 billion.
This pending proliferation of different coherent solutions opens up the question of which option is best suited for which application. For short-reach, point-to-point links between data centers across a metro, use of pluggable coherent transceivers directly in packet switch ports provides cost savings by eliminating the need for paired sets of short-reach optics connecting the packet switch ports to WDM transponder systems, and the extra cost, space and power of the latter.
High-performance coherent optics having a "headline capacity" of 800Gbit/s, which requires complex modulation such as 64QAM, will be limited to reaches of around 100km, and wavelength speed rapidly reduces, to 600Gbit/s, 500Gbit/s, or less when used on longer links. Compared to using pluggable 400G ZR transceivers for short-reach applications, chassis-based 800Gbit/s coherent optics will require more cost, space and power, calling into question the near-term value of 800Gbit/s per wave as a useful feature.
The value of high-performance coherent optics will come from the ability to upgrade long-haul datacenter interconnections and national-scale packet networks to 400GE speeds at scale, over greater distances than is possible with 400G ZR or ZR+. Next-generation high-performance coherent optics will enable 400Gbit/s transport across virtually any distance, including long-haul and subsea links spanning thousands of kilometers.
Network upgrades supporting 400Gbit/s over a single wavelength will be more spectrally efficient than prior generations of coherent optics, which require bonding of multiple, lower speed wavelengths to transport 400GE across long-distance links. Combined with use of optical line systems operating in the C+L bands of the fiber spectrum, network operators will be able to efficiently transport high bandwidth services, while more than doubling total fiber capacity of existing fiber networks.
As application-optimized coherent solutions and vendors proliferate and diversify, network operators will pick the best products for the job. For vendors, success will come from developing the proper balance in product capabilities and optimizing the DSPs, optics and packaging implementations accordingly. Meanwhile, the race to vertically integrate key aspects of coherent technology will continue to make the coherent optics ecosystem a space to watch for all of 2020.
— Serge Melle works in sales enablement for Nokia IP-Optical Networks. You can visit his LinkedIn profile here.